Indira Awas Yojana and Housing Adequacy: An Evaluation

1

ASARC Working Paper 2019/05

Indira Awas Yojana and Housing Adequacy: An Evaluation using Propensity Score Matching

Amarendra Sharma1

Abstract

Indira Awas Yojana (IAY) was a public housing assistance program of the Government of India with its primary purpose being the provision of housing to the homeless rural poor households. The efficacy of this program has not been rigorously evaluated in the literature and thus this paper is an attempt in that direction. This paper, by using IHDS-II, a nationally representative sample from India, and relying on the quasi-experimental technique of Propensity Score Matching, concludes that the IAY has been moderately successful in meeting its goals in term of housing characteristics such as pucca house, pit toilet, smokeless chulha, hand pump water; but, it fails to reach its stated goals in terms of external housing adequacy such as the presence of excrement/human waste and stagnant water in the vicinity of house and the access to village health facility and electrification.

JEL Classifications: H53, I38, O18, R2

Keywords: Indira Awas Yojana, Public Housing Program, Propensity Score Matching, Impact Evaluation, India.

1 Associate Professor of Economics, Elmira College, NY, USA. Email: [email protected]. Also affiliated with the School of International Service, American University, Washington DC, USA. Email: [email protected].

2

1.0 Introduction

The access to decent housing is included in the United Nation’s Sustainable Development Goals

(Goal 11). It sees housing as an essential human need which is needed for survival and basic human

decency. Housing is also linked to ownership, sense of identity, community building, and

promotion of self-esteem and confidence. Since housing supports livelihoods and promotes social

integration, it is deemed critical for an individual’s social and economic development. Moreover,

home ownership is viewed as a form of cultural expression and a symbol of social standing.

The government of India acknowledged the role of housing in its strategy of poverty alleviation,

and subsequently, launched several public housing assistance programs over the years beginning

with Community Development Movement (CDM), a Village Housing Programme (VHP) in 1957,

which provided loans to individuals and cooperatives of up to Rs. 5,000 per unit. The most

ambitious public housing program called Indira Awas Yojana (IAY) was launched in June 1985

by the Government of India, as a sub-scheme of the National Rural Employment Program (NREP),

with an objective of removing poverty through the provision of housing to the rural households.

Indira Awas Yojana is a publicly funded program that aims to provide homes to homeless rural

poor households; and, also to those who live in dilapidated and kutcha homes. This program also

provides loans to the landless poor for the sole purpose of buying land to construct house. The

program intends to provide access to housing to SC/STs, female headed households, households

with disability, and marginalized households. The program stipulates that every house under IAY

should include a toilet, soak pit and compost pit, smokeless chulhas (exempted if the households

have an LPG /biogas connection). A provision for the roof water harvesting system is also allowed

3

if the local conditions are appropriate. The program also encourages households to construct a

bathroom.2

This paper aims to evaluate the effectiveness of Indira Awas Yojana since there is no systematic

and rigorous study that addresses this issue in the academic literature. This paper relies on a quasi-

experimental statistical tool of propensity score matching to perform impact evaluation. The non-

availability of data from a randomized control trial makes propensity score matching a suitable

strategy for the available cross-sectional survey data. The data for this study is obtained from the

India Human Development Survey-II (IHDS-II), which was collected by NCAER and University

of Maryland. This data covers more than 42,000 households from across India and is representative

of the whole country.

This paper aims to compare the households who availed the housing loans under the IAY with

those who did not but were comparable in every other observable respect. The outcome variables

aim to include those measures that reflect the objectives specified under the IAY such as housing

quality, linkage with environment and infrastructure, which collectively can be termed as housing

adequacy that we further classify into two categories, viz, internal and external.

2.0 Related Literature

Unfortunately, the literature is sparse in its coverage of IAY. There are a few studies that have

tried to evaluate the IAY with regards to the number of homeless households who benefited under

this plan (see, for example, Venkateswarlu, 2017; Shivanna & Kadam, 2017a; Shivanna & Kadam,

2017b). However, these studies, lack the rigors required of an impact evaluation of a public housing

assistance program. There are two government level evaluations of IAY that we are aware of. The

first one is a study carried out by the Department of Economic and Statistical Analysis,

2 The above description is obtained from the IAY Guidelines (http://iay.nic.in/netiay/IAY%20revised%20guidelines%20july%202013.pdf).

4

Government of Haryana, which focuses on several aspects of IAY’s impact on beneficiaries

including whether the objectives of the scheme are being achieved, whether the beneficiaries are

receiving the allocated funds, whether the funds are disbursed in a smooth and timely fashion,

whether the funds are being utilized in the proper way by the beneficiaries, and whether the houses

are built as per the guidelines and in a timely fashion (See, Bishnoi, 2012). The last aspect is

relevant to this study since it evaluates whether the built houses under the IAY met the guidelines

of the program. Bishnoi (2012) using a sample of 160 households, who participated in the program

from Haryana, concludes that the program met its goal of providing pucca houses to all

participants. It also concludes that the program was successful in terms of the provisioning of

smokeless chulha, sanitation and ventilation facilities. However, it also found that the program

was not successful in terms of the provision of drinking water. Even though, the effort on the part

of the Government of Haryana to assess the efficacy of IAY is laudable, it is naïve and rudimentary

in its approach since the analysis lacks the statistical rigor required of an evaluation study, as will

be made clear subsequently in the next section.

The second government level evaluation is by the Comptroller and Auditor General (CAG) of

India, which audited the IAY from 2008 to 2013 in 168 districts across 27 states and 4 union

territories. The CAG report is critical of the program and found several irregularities in its

implementation, notably in the identification and selection of the beneficiaries and the construction

of house and the quality of house.3

Unfortunately, we could not find a single study that attempts to evaluate the impact of IAY on

housing adequacy using a nationally representative sample and apposite methodology.

3https://cag.gov.in/sites/default/files/audit_report_files/Union_Performance_Indira_Awaas_Yojana%20_37_2014_chapter_3_exe-sum.pdf

5

This study assumes importance because merely providing financial and technical assistance for

home construction and subsequently counting the number of homes constructed and then claiming

that the program met its goal, could be misleading. Thus, it is imperative that the program should

be evaluated on the basis of quality and adequacy of constructed homes, as per the objectives laid

down in the IAY program.

3.0 Methodology

The chief econometric concern that we face in estimation is the issue of self-selection, which is

likely to bias the estimates. It is plausible to think that there are some intrinsic differences between

households that choose to participate and those that do not, and if we do not observe these

characteristics, it is possible that the IAY specific housing adequacy outcomes are not only driven

by the participation but also by the unobservable factors. In this scenario, the true casual impact of

participation in IAY on housing adequacy outcomes is hard to estimate through standard OLS

regression. In other words, the estimate is likely to be biased.

To assess the true causal impact of participation in IAY on housing adequacy outcomes, we will

need to know what would have been the housing adequacy outcomes of a household that chose to

participate in IAY had it not participated in IAY. Unfortunately, the same household cannot be at

the same time in the program and not in the program. However, there are tools that allow us to

create reasonable statistical counterfactuals to the treatment group.

The gold standard in impact evaluation is the Randomized Control Trial (RCT), whereby a

randomly selected group of households are chosen to participate in the program and another group

with similar characteristics serves as control without receiving participation. However, such an

6

experiment is impractical in many circumstances due to a set of complex issues related to

participation, and IAY is no exception in that regard. That leaves us with quasi-experimental

techniques, such as Propensity Score Matching (PSM), which allow us to mimic experiments, in

the sense, that we can create statistical counterfactuals to the treatment group and derive

meaningful results without conducting the experiment.

The PSM strategy has been documented to reduce selection bias (see, for example, Dehejia &

Wahba, 1998; Hong & Raudenbush, 2005; Shadish et al., 2002). Propensity score matching, if

appropriately used, should yield relatively unbiased estimates of IAY’s effects on housing

adequacy measures. Results obtained from quasi-experiments using propensity score matching

methods can closely approximate those obtained from randomized control trials (Becker & Ichino,

2002). For example, Luellen, Shadish, and Clark (2005) contrasted findings from two quasi-

experiments using propensity score matching to those obtained from two true experiments. Use of

propensity score matching reduced selection bias by 73–90%. The mean differences obtained from

the propensity score analyses and those obtained using randomization differed by only .09 and .20

of a point on each study’s particular outcome measure. Because of its ability to greatly reduce

selection bias, propensity score matching is increasingly being utilized in the fields of policy

evaluation and economics (see, for example, Harknett, 2006; Jones, D’Agostino, Gondolf, &

Heckert, 2004; Czajka, Hirabayashi, Little, & Rubin, 1992; Lechner, 2002; Bryson at al., 2002;

Levine & Painter, 2003; List et al., 2003; O'Keefe, 2004, Jalan & Ravallion 2003; Trujillo, Portillo,

& Vernon, 2005; Lynch, Gray, & Geoghegan, 2007; Mendola, 2007; Pufahl & Weiss, 2009; Oh

et al., 2009; Cox-Edwards & Rodríguez-Oreggia, 2009; Mensah, Oppong, & Schmidt, 2010;

Becerril & Abdulai, 2010; Wamser, 2014; Chiputwa, Spielman, & Qaim, 2015).

7

Under the PSM strategy, the selection into treatment group is based on observable characteristics.

Every household is assigned a propensity score based on observable characteristics that determine

participation in IAY in the first stage. We use the following observable characteristics: an indicator

of whether household lives in the rural area, an indicator of whether the household is below the

official poverty level, an index of household assets, an indicator of whether the household

owns/cultivates agricultural land, the educational attainments of male and female household heads,

household size, an indicator of whether the household belongs to SC/ST social group, religious

affiliation of the household, an indicator of whether the household has acquaintance with elected

officials, measures of household exposure to media, household size, number of adult males and

females in the household, the age of male and female household heads, an indicator of whether the

household has acquaintances with local elected politicians, and the district and state of the

household.

In the second stage, several matching algorithms could be employed to compare the mean

outcomes between the treatment and control groups. We primarily rely on Radius Matching with

a caliper of 0.01, which has been used in other impact evaluation studies as well (see for example

Kumar & Volmer, 2013). However, for the sake of robustness of our result, we will also report the

results from Nearest Neighbor Matching. The matching algorithms are briefly discussed below.4

The Nearest-Neighbor Matching (NNM): Under this algorithm each treatment unit is matched

to the comparison unit with the closest propensity score. Matching can be done with or without

replacement. Matching with replacement means that the same nonparticipant can be used as a

match for different participants. We use the five nearest neighbors matching with replacement.

4 The following discussion is borrowed from Khandker et al (2010).

8

Radius Matching (RM): One problem with NN matching is that the difference in propensity

scores for a participant and its closest nonparticipant neighbor may still be very high. This situation

results in poor matches and can be avoided by imposing a threshold or “tolerance” on the maximum

propensity score distance (caliper). This procedure therefore involves matching with replacement,

only among propensity scores within a certain range.

Propensity Score-based weighted regression

For robustness check, we rely on the estimation of a weighted multiple regression model, which

uses the propensity scores as sampling weights. It has been suggested that this weighting

procedure, which relies on the propensity score, balances the distribution of covariates, and yields

fully efficient estimates (see, for example, Rosenbaum, 1987; Hirano and Imbens, 2001; Hirano et

al., 2003). We weight the treatment and control groups by their respective propensity scores and

obtain covariate distribution, which is similar for the two groups. The inverse of the propensity

score, serves as the weight for the treatment group; while the inverse of one minus the propensity

score, serves as the weight for the control group.

4. Data and Descriptive Statistics

This study relies on data from the second round of India Human Development Survey (IHDS-II),

which was administered in 2011-12. According to the IHDS-II descriptions, it is a nationally

representative, multi-topic survey of 42,152 households in 384 districts, 1420 villages and 1042

urban neighborhoods across India. The survey covers almost all states and union territories of

India. The survey relied on a couple of one-hour interviews in each household to elicit information

on topics such as covered health, education, employment, economic status, marriage, fertility,

9

gender relations, and social capital. IHDS was jointly executed by researchers from the University

of Maryland and the National Council of Applied Economic Research (NCAER), New Delhi

(Desai et al, 2015).

4.1 Outcome Variables

In this paper, our objective is to evaluate the effectiveness of IAY with regards to internal and

external housing adequacy by considering the outcomes that the program aims to deliver. The

program stipulates certain housing standards to be met to qualify for the loan. First, the house built

under this program must be ‘pucca ‘, implying that it should be able to endure normal wear and

tear due to usage and natural forces including climatic conditions, with reasonable maintenance,

for at least 30 years. It should have roof of permanent material and its walls should be capable of

withstanding local climatic conditions and need to be plastered only when the outer surface of the

walls is erodible. Hence, we consider three such housing adequacy outcome variables; namely,

whether the house has pucca walls, pucca roof, and pucca floor. Approximately, 73% of the

sampled households have pucca walls, 85% have pucca roof, and 66% have pucca floor.

Approximately, 57% of the sampled households have houses with pucca walls, pucca roof, and

pucca floor. Second, every house should include a toilet, soak pit and compost pit. To assess toilet

specific outcome, we consider four variables that indicate whether the household has no access to

toilet (open field defecation), pit toilet, septic toilet, and flush toilet. Approximately, 44.5% of the

sampled households have no access to toilets, implying that they defecate in open fields.

Approximately, 14.9% of the sampled households have access to pit toilets at home. Whereas,

30.4% of the sampled households have access to septic toilets and 9.75% to the flush toilets at

home. Unfortunately, the IHDS-II did not contain any information on soak pit and compost pit,

and therefore, we are unable to include these outcomes in this study. Third, the program

10

participation requires the household to construct smokeless chulha, which however can be

exempted for households in possession of an LPG /biogas connection. We consider four chulha

specific variables to assess this goal. These variables are: open fire chulha, traditional chulha,

improved chulha, and non-biomass chulha. Approximately, 16.1% households cook food in open

fire chulhas, 39.9% in traditional chulhas, 7% in improved chulhas, and 36.3% in non-biomass

chulhas. Fourth, the IAY also provides loans for acquiring land sites for the sole purpose of

constructing house. The program stipulates that while selecting land, it should be ensured that it is

fit for construction of houses especially in terms of physical connectivity, power connectivity,

availability of drinking water, access to public institutions, etc. We assess these goals by

considering whether the house has access to clean water, tube well, hand pump, open well, covered

well, piped water, electricity, absence of excrement/stagnant water surrounding the house, and

closeness of health facility. Approximately, 92.8% of the sampled households have access to clean

water. Approximately, 11% of the sampled households rely on tube wells for water; while, 26.4%

rely on hand pumps. Approximately, 1.5% of the households rely on covered wells and 8.3% on

open wells for water. Approximately, 48.2% of the sampled households rely on piped water. As

far as the external housing adequacy is concerned; approximately, 87.3% of the sampled

households have electricity at home. Approximately, 21.8% of the households have the presence

of excrement/human waste surrounding their homes. Whereas, approximately, 15.3% of the

sampled households have the presence of stagnant water near their homes. Approximately, 54.4%

of the sampled households have a health facility in their villages.

4.2 Treatment Variable

For the treatment variable, we utilize the information from the following question that the IHDS-

II asked survey participants: During the past five years did you or any other member of your

11

household participate in/ benefit from social insurance schemes (government or private)? One of

the options to this question was Indira Awas Yojana. Out of 41,975 households who responded to

this survey question, approximately 5.13% of the households had participated/benefited from IAY.

4.3 What Determines Selection into Program?

The design of IAY elicits participation from households with certain characteristics, as it is

voluntary in nature, even though it seeks to target poor homeless families in rural areas. Thus, it is

essential to understand what determines participation into public housing program. The stipulation

of IAY suggests that households without homes or who live in kutcha/dilapidated homes and are

below the official poverty level and belong to SC/ST, female headed households, and households

with disability, and marginalized households are eligible for participation in the program. Since

the program relies on households self-selecting themselves into participation, it is worth learning

what determines their participation. The observed household characteristics that are likely to

influence participation into IAY include whether the household belongs to SC/ST social group, is

below poverty level, household assets, tills/own agricultural land, religion, size of the household,

educational attainment of male and female household heads, an indicator for residing in the rural

area, a measure of exposure to media, a measure of acquaintance with local elected politicians,

and the age of the male and female household heads. We also control for household’s district and

state, as IAY participation rules are not uniformly implemented across all states.

Approximately, 34% of the surveyed households are in the possession of BPL-based ration cards.

Almost 65.4% of the surveyed households live in the rural areas. The IHDS-II constructs an index

to measure household assets based on the year 2005 assets possessed by the household. This index

ranges from 0-30. The mean score of the household assets based on this index is 14.85.

Approximately, 44.5 percent of the surveyed households owned or cultivated agricultural lands.

12

The average years of schooling for the male head of the household in this sample is 7.8 years with

standard deviation of 5. The maximum years of schooling is 16 years, and the minimum is zero

years. The average years of schooling for the female head of the household in this sample is 5.6

years with standard deviation of 5.2. The maximum years of schooling is 16 years, and the

minimum is zero years. The average size of the household is 4.85 persons with standard deviation

of 2.3. The largest household comprised of 33 people and the smallest consisting of only one

person. The average number of adult males in a household is 1.42 with standard deviation of 0.87.

The highest number of adult males observed in a household is 9 and the smallest being zero. The

average number of adult females in a household is 1.49 with standard deviation of 0.78. The

highest number of adult females observed in a household is 9 and the smallest being zero.

Approximately, 30 percent of the surveyed households belonged to the SC/ST social group. The

average age of the male household head is 49.15 years with standard deviation of 13.5 years. The

highest age for this variable is 99 years and the lowest age is 15 years. The average age of the

female household head is 44.61 years with standard deviation of 13.06 years. The highest age for

this variable is 99 years and the lowest age is 7 years. Approximately, 3.8% of the surveyed

households had membership in a political party and 29% of the households attended political

meetings. On a scale of 1-3, to indicate a household’s confidence in the local panchayat’s ability

to implement public projects, the average score is 1.91 with standard deviation of 0.68.

Approximately, 13.2% of the surveyed households had acquaintances with elected politicians

outside of their community/caste.

5.0 Results

5.1 Naive Result

13

Table-1 presents the estimation result of linear probability model (LPM) where we control for a

host of influencers on the outcome variables.5 The estimation results suggest that the IAY

households are 6.4% more likely to live in houses with pucca walls than the non-participants. This

result is statistically significant at the 1% level. Also, the homes of IAY households are 2.1%

more likely to have pucca roofs relative to the non-participants. This result is statistically

significant at the 5% level. Whereas, there is no statistically significant difference between IAY

households and non-IAY households in terms of living in houses with pucca floors and pucca

houses (pucca wall, pucca roof, and pucca floor).

With regards to the non-availability of toilet at home and the flush toilet at home, there are no

statistically significant differences between the IAY participants and non-participants. The IAY

participants are 1.6% more likely to have pit toilet at home compared to the non-participants. This

result is statistically significant at the 10% level. However, the IAY participants are 1.7% less

likely to have septic toilets at home relative to the non-participants, and this result is statistically

significant at the 5% level.

The IAY participants are 2.1% more likely to use the open fire chulha relative to the non-

participants, and the result is statistically significant at the 5% level. Whereas, the non-participants

are 3.1% more likely to use traditional chulha than the participants. The result is statistically

significant at the 1% level. We did not find any statistically significant difference between the two

groups in terms of using either improved chulha or non-biomass chulha.

5 An appropriate model to use would be a logit or probit. However, the impact evaluation literature typically reports the results from the LPM, in spite of its shortcomings in dealing with the binary dependent variables. The reason being that the estimates under LPM can be directly and conveniently compared with the results of quasi-experimental models, such as the PSM based models. It’s worth noting that the primary conclusions of the paper are not based on this model.

14

When it comes to water, we did not find any statistically significant difference in terms of either

the availability of drinking water or the access to piped water between the two groups of

households. However, the IAY households are 2.1%, 1.9%, and 0.9% less likely to access water

from tube well, covered well, and open well relative to the non-participants. These results are

statistically significant at the 1% level. Whereas, the IAY participants are 6.7% more likely to rely

on hand pump water compared to the non-participants, and this result is statistically significant at

the 1% level.

The IAY participants are 3.1% less likely to have electricity in their homes relative to the non-

participants. This result is statistically significant at the 1% level. Whereas, the IAY participants

are 2.6% more likely to have the presence of stagnant water near their homes relative to the non-

beneficiary homes. This result is statistically significant at the 1% level. Moreover, we did not find

any statistically significant difference in either the presence of excrement/human waste

surrounding homes or the access to village health facility between the two groups.

5.2 Propensity Score Estimation

Table 2 column 6 presents the result of logit model of program participation where the dependent

variable is participation in IAY. It is evident that most of the observed covariates are statistically

significant in explaining participation. Below poverty line, rural, SC/ST, size of the family, age of

the male household head, attend public meeting, and confidence in local panchayat positively

predict participation in IAY. Whereas, household assets, educational attainment of male and

female household heads, acquaintance with elected politician, age of the female household head

negatively predict participation in IAY.

5.3 Post Matching Quality

15

There are several tests available to assess how good the quality of a match is for the covariates in

the post-matching scenario. A good matching is an indicator of a covariate being similar in both

pre- and post-matching cases. One can look at the mean difference in pre- and post-matching

scenarios for each covariate and the resulting t-values. If the t-statistic is not significant, then we

can conclude that the matching is good. Our results suggest that excellent matching is achieved in

all covariate cases. Additionally, there are other summary measures to assess the quality of

matching such as Psuedo-R2 statistic and Likelihood-Ratio test. A low value of Psuedo-R2 and a

high value of the Likelihood-Ratio test is a sign of good matching. The results presented in the

table for radius matching overwhelmingly suggest that we have an excellent matching and were

able to reduce the observed selection bias.

In order to assess, if we have sufficient overlap in the propensity scores between the treated and

non-treated groups, since the matching process includes only those observations that are on

common support; we also provide a graph of the density distribution of the propensity scores. The

visual inspection of the density distribution of the propensity score suggests that we have sufficient

overlap between the treated and the control groups, and thus satisfies the overlap condition of the

matching process.

5.4 PSM Based Estimation Results

Table-4 presents the estimation results of PSM using radius matching with a caliper of 0.01. The

estimation results suggest that the IAY households are 8.4% more likely to live in houses with

pucca walls than the non-participants. This result is statistically significant at the 1% level. Also,

the homes of IAY households are 3% more likely to have pucca roofs relative to the non-

participants. This result is statistically significant at the 1% level. Whereas, IAY households are

16

2.9% more likely to live in houses with pucca floors than the non-participants, and the result is

statistically significant at the 5% level.

With regards to the non-availability of toilet at home and the septic toilet at home, there are no

statistically significant differences between the IAY participants and non-participants. The IAY

participants are 2.8% more likely to have pit toilet at home compared to the non-participants. This

result is statistically significant at the 1% level. However, the IAY participants are 1.4% less likely

to have flush toilets at home relative to the non-participants, and this result is statistically

significant at the 1% level.

The IAY participants are 2% more likely to use the open fire chulha relative to the non-participants,

and the result is statistically significant at the 10% level. Whereas, the non-participants are 2.4%

more likely to use traditional chulha than the participants. The result is statistically significant at

the 10% level. The IAY participants are 1.4% more likely to use the improved chulha relative to

the non-participants and the result is statistically significant at 10% level. We did not find any

statistically significant difference between the two groups in terms of using the non-biomass

chulha.

With regards to water, we did not find any statistically significant difference in terms of either the

availability of drinking water or the access to piped water between the two groups of households.

However, the IAY households are 2.2%, 0.7%, and 2.2% less likely to access water from tube

well, covered well, and open well relative to the non-participants. These results are statistically

significant at the 1% level. Whereas, the IAY participants are 6.3% more likely to rely on hand

pump water compared to the non-participants, and this result is statistically significant at the 1%

level.

17

The IAY participants are 2.4% less likely to have electricity in their homes relative to the non-

participants. This result is statistically significant at the 5% level. Whereas, the IAY participants

are 3.1% more likely to have the presence of stagnant water near their homes relative to the non-

beneficiary homes. This result is statistically significant at the 1% level. Moreover, we did not find

any statistically significant difference in either the presence of excrement/human waste

surrounding homes or the access to village health facility between the two groups.

5.5 Sensitivity Analysis

Given that the PSM only accounts for the biases due to the unobserved characteristics, a reliance

on it does not imply that the biases due to the unobserved variables will be eliminated. Thus, it is

plausible to think that there exist some unseen biases that could potentially influence our estimates.

One such example might be that the households who participated in IAY are more motivated, and

therefore are more likely to take measures to improve the quality of their homes, even if they didn’t

participate in the program. This leads us to believe that our estimates could be biased since the

extent of their motivations could not be observed. The estimation of treatment effects with

matching methods crucially depends on the assumption of conditional independence, meaning that

the treatment and control groups do not differ on unobservable attributes that influence the

selection into treatment and outcome variables. Failing this assumption, one is likely to obtain non

robust estimates from the matching methods (Rosenbaum, 2002). The selection bias is hard to

quantify because we are using the non-experimental data. Thus, we employ the Rosenbaum’s

(2002) method to ascertain the extent of unobservable variables influence such that the estimated

treatment effects become invalid. We present the results from the sensitivity analysis in table 5.

First, we analyze the Q _mh and P_mh statistics. These statistics suggest that the PSM estimates

are not sensitive to the selection bias due to the unobserved characteristics, and the estimated

18

treatment effects remain statistically significant even in the presence of large unobserved biases.

Next, we examine Q_mh+ and P_mh+ statistics for the positive unobserved selection bias. This is

likely to occur when IAY households may have higher rates of adequate housing, which results in

upward bias and needs to be adjusted downwardly. On the other hand, the negative unobserved

selection bias implies that those households who participate in IAY are also more likely to have

lower rates of housing adequacy, thereby underestimating the treatment effects, which therefore

needs to be adjusted upwardly. The results suggest that even after allowing for a significant amount

of positive or negative selection on unobservable characteristics, the PSM estimates remain

statistically significant.6

6.0 Inverse Probability Weighted Regression

Table-6 presents the estimation results of inverse probability weighted regression. These

estimations fully support the conclusions obtained under the PSM model. The participants in IAY

are 6.6% more likely to have pucca walls, 3% more likely to have pucca roof, and 1% more likely

to have pucca floor relative to the non-participants. These estimates are statistically significant at

the 1% level. The IAY participants are 1.3% less likely to defecate in open fields (no toilets), 4%

more likely to have pit toilets, 0.6% more likely to have septic toilets, and 3.2% less likely to have

flush toilets relative to the non-participants. These estimates are statistically significant at the 1%

level. With regards to the provision of chulha, the IAY participants are 2.3% more likely to have

open fire chulha, 1.5% less likely to have traditional chulha, 1.2% more likely to have improved

chulha, and 1.9% less likely to have non-biomass chulha relative to the non-participants. These

estimates are statistically significant at the 1% level. When we look at the availability of water, the

6 The sensitivity analysis used in this paper is borrowed from Kumar & Vollmer (2013).

19

IAY participating households are 1.5% more likely to have adequate availability of water than the

non-participating households, and the result is statistically significant at the 1% level. Moreover,

the proportion of households that rely on tube well water is 2% less for the IAY participating

households relative to the non-participants, and the result is statistically significant at the 1% level.

Whereas, IAY households are 5.8% more likely to rely on handpump water relative to the non-

participants, and the result is statistically significant at the 1% level. Additionally, IAY households

are 1.4% and 1.2% less likely to rely on open and covered well water, respectively, relative to the

non-participating households. These estimates are statistically significant at 1% level. We did not

find any statistically significant difference in the reliance on piped water between the two groups.

IAY households are 0.7 percent less likely to have access to electricity and 1.4% more likely to

have the prevalence of excrement surrounding their homes relative to the non-participating

households. These results are statistically significant at the 1% level. The IAY households are

6.3% more likely to have stagnant water near their homes and 2.9% more likely to have access to

the village health facility relative to the non-participating households. These results are statistically

significant at 1% level.

7.0 Alternative PSM Models

Table 7 presents the estimation results of PSM model using the Nearest-Neighbor matching

algorithm. The results, more-or-less, confirm the conclusions drawn under radius matching.

8.0 Discussion and Limitations

First, we discuss the results concerning the internal housing adequacy outcomes. Our results, based

on all model specifications, unequivocally suggest that a greater proportion of IAY beneficiaries

(treated group) live in homes with pucca walls than the non-beneficiaries (control group). All

20

model specifications except one, also suggest that a greater proportion of the treated households

live in homes with pucca roof than the control households. Although, greater proportion of the

treated households live in homes with pucca floors relative to the control households, the result is

not statistically significant in some model specifications. Overall, it can be concluded that the

participation in IAY resulted in homes with pucca walls, pucca roof, and pucca floors, making the

program a success on this outcome measure.

The results concerning the external aspects of housing adequacy as measured by electricity

connection, presence of excrement/human waste matter surrounding the house, presence of

stagnant water near the house, and presence of village health facility do not show any positive

impact on the participants. One explanation for this could be found in the fact that many of the

participants already owned the land and availed the loan facility to construct new house or improve

the existing house. If these land sites lacked external housing adequacy, the participation in IAY

is not likely to substantially improve these aspects, as it is not explicitly required to participate. A

positive outcome could only be expected for those participants who took the loan to purchase land

to construct house, since the participation is contingent on meeting the required guidelines

pertaining to the external housing adequacy. Another possibility is that the funding for the external

housing adequacy might have come from some other programs, such as Jawahar Rojgar Yojana or

NREGA, which have their own goals and pace of implementation, and therefore external housing

adequacy might not manifest in a prompt fashion with the objectives of IAY. The third possibility

is poor implementation of the program and lack of oversight and accountability.

Since the paper relies on propensity score matching, which relies on matching between the

treatment and control groups taking place solely on observed characteristics, it is possible that

some unobserved characteristics might have systematically influenced the outcome variables.

21

Unfortunately, the PSM technique cannot address this possibility and maintains a strong

assumption that the unobserved characteristics did not influence the outcome variables.

The results of this paper based on PSM and IPW regression do point towards the success of the

IAY in terms of meeting some of the program objectives. However, in spite of, our carefully

executed statistical exercises to establish causal links between participation in IAY and various

housing adequacy measures considered in this paper, we refrain from claiming the causal links,

due to the strong assumptions required for such a claim under the PSM approach. However, it

should not go without notice that this study shows a very strong statistical association between the

two measures.

9.0 Conclusion

Indira Awas Yojana, a public housing assistance program of the government of India, started in

1985 as a part of the mission to eradicate poverty, and subsequently becoming a full-fledged

program in 1996 and now rechristened as the Pradhan Mantri Awas Yojana, aims to provide

housing to the rural homeless poor households, SC/STs, other marginalized sections of the society,

disabled and female-headed households.

This study aims to evaluate the effectiveness of IAY in terms of the outcome goals laid out in the

program with regards to internal and external housing adequacy by using a nationally

representative sample consisting of more than 42,000 households from across India, and

employing the quasi-experimental technique of propensity score matching.

The results of this paper suggest that IAY has been a moderately successful program especially

with regards to the internal housing adequacy measures such as the pucca structure of house,

provision of toilet, provision of smokeless chulha, availability of drinking water. However, the

22

program cannot be characterized as success in terms of the external housing adequacy related

measures, such as electrification, presence of excrement/human waste and stagnant water, and the

availability of village health facility. Although, it should be noted that the shortcomings in terms

of external housing adequacy measures cannot solely be attributed to IAY, since other government

programs are also tasked with the improvement of rural infrastructure.

Based on the findings of this paper, we recommend that the government of India should continue

with this program, but more emphasis should be given to a better implementation in terms of

accomplishing the external housing adequacy outcomes in addition to making sure that the

program is able to yield hundred-percent of the internal housing adequacy outcome measures.

References:

Becker, Sascha O., and Andrea Ichino. "Estimation of average treatment effects based on propensity scores." The stata journal 2, no. 4 (2002): 358-377.

Bishnoi, R. K. "Evaluation study of Indira Awaas Yojana." An Evaluation Report issued by Department of Economic and Statistical Analysis, Government of Haryana 145 (2012): 1-49.

Becerril, Javier, and Awudu Abdulai. "The impact of improved maize varieties on poverty in Mexico: a propensity score-matching approach." World development 38, no. 7 (2010): 1024-1035.

Bryson, Alex, Richard Dorsett, and Susan Purdon. "The use of propensity score matching in the evaluation of active labour market policies." (2002).

Chiputwa, Brian, David J. Spielman, and Matin Qaim. "Food standards, certification, and poverty among coffee farmers in Uganda." World Development 66 (2015): 400-412.

Cox-Edwards, Alejandra, and Eduardo Rodríguez-Oreggia. "Remittances and labor force participation in Mexico: an analysis using propensity score matching." World Development 37, no. 5 (2009): 1004-1014.

Czajka, John L., Sharon M. Hirabayashi, Roderick JA Little, and Donald B. Rubin. "Projecting from advance data using propensity modeling: An application to income and tax statistics." Journal of Business & Economic Statistics 10, no. 2 (1992): 117-131.

Dehejia, Rajeev H., and Sadek Wahba. "Propensity score matching methods for non-experimental casual studies." (1998).

Desai, Sonalde, Amaresh Dubey, and Reeve Vanneman. India Human Development Survey-II (IHDS-II) [Computer file]. University of Maryland and National Council of Applied Economic

23

Research, New Delhi [producers], 2015. Ann Arbor, MI: Inter-university Consortium for Political and Social Research [distributor].

Harknett, Kristen. "The relationship between private safety nets and economic outcomes among single mothers." Journal of Marriage and Family 68, no. 1 (2006): 172-191.

Hong, Guanglei, and Stephen W. Raudenbush. "Effects of kindergarten retention policy on children’s cognitive growth in reading and mathematics." Educational Evaluation and Policy Analysis 27, no. 3 (2005): 205-224.

Indira Awas Yojana: Guidelines. Accessed at: http://iay.nic.in/netiay/IAY%20revised%20guidelines%20july%202013.pdf

Jalan, Jyotsna, and Martin Ravallion. "Estimating the benefit incidence of an antipoverty program by propensity-score matching." Journal of Business & Economic Statistics 21, no. 1 (2003): 19-30.

Jones, Alison Snow, Ralph B. D’Agostino Jr, Edward W. Gondolf, and Alex Heckert. "Assessing the effect of batterer program completion on reassault using propensity scores." Journal of Interpersonal Violence 19, no. 9 (2004): 1002-1020.

Khandker, Shahidur R., Gayatri B. Koolwal, and Hussain A. Samad.Handbook on impact evaluation: quantitative methods and practices. World Bank Publications, 2010.

Kumar, Santosh, and Sebastian Vollmer. "Does access to improved sanitation reduce childhood diarrhea in rural India?" Health Economics 22, no. 4 (2013): 410-427.

Lechner, Michael. "Program heterogeneity and propensity score matching: An application to the evaluation of active labor market policies." The review of economics and statistics 84, no. 2 (2002): 205-220.

Levine, David I., and Gary Painter. "The schooling costs of teenage out-of-wedlock childbearing: Analysis with a within-school propensity-score-matching estimator." The Review of Economics and Statistics 85, no. 4 (2003): 884-900.

List, John A., Daniel L. Millimet, Per G. Fredriksson, and W. Warren McHone. "Effects of environmental regulations on manufacturing plant births: evidence from a propensity score matching estimator." The Review of Economics and Statistics85, no. 4 (2003): 944-952.

Luellen, Jason K., William R. Shadish, and M. H. Clark. "Propensity scores: An introduction and experimental test." Evaluation Review 29, no. 6 (2005): 530-558.

Lynch, Lori, Wayne Gray, and Jacqueline Geoghegan. "Are farmland preservation program easement restrictions capitalized into farmland prices? What can a propensity score matching analysis tell us?" Review of agricultural economics29, no. 3 (2007): 502-509.

Mendola, Mariapia. "Agricultural technology adoption and poverty reduction: A propensity-score matching analysis for rural Bangladesh." Food policy 32, no. 3 (2007): 372-393.

24

Mensah, Joseph, Joseph R. Oppong, and Christoph M. Schmidt. "Ghana's National Health Insurance Scheme in the context of the health MDGs: An empirical evaluation using propensity score matching." Health economics 19, no. S1 (2010): 95-106.

Oh, Inha, Jeong-Dong Lee, Almas Heshmati, and Gyoung-Gyu Choi. "Evaluation of credit guarantee policy using propensity score matching." Small Business Economics 33, no. 3 (2009): 335-351.

O'Keefe, Suzanne. "Job creation in California's enterprise zones: a comparison using a propensity score matching model." Journal of Urban Economics 55, no. 1 (2004): 131-150.

Pufahl, Andrea, and Christoph R. Weiss. "Evaluating the effects of farm programmes: results from propensity score matching." European Review of Agricultural Economics 36, no. 1 (2009): 79-101.

Rosenbaum, Paul R., and Donald B. Rubin. "The central role of the propensity score in observational studies for causal effects." Biometrika 70, no. 1 (1983): 41-55.

Rosenbaum, Paul R. "Observational studies." In Observational studies, pp. 1-17. Springer New York, 2002.

Shivanna, T., and R. N. Kadam. "Performance Appraisal of Indira Awas Yojana in Chitradurga District, Karnataka." (2017).

Shivanna, T., and R. N. Kadam. "An Evaluation of Indira Awas Yojana in Challakere Taluk of Chitradurga District, Karnataka." (2017).

Venkateswarlu, Rapolu. "An Analysis of Indira Awas Yojana Scheme in Kodad Constituency of Suryapet District, Telangana State." (2017).

Shadish, William R. "Revisiting field experimentation: field notes for the future." Psychological methods 7, no. 1 (2002): 3.

Trujillo, Antonio J., Jorge E. Portillo, and John A. Vernon. "The impact of subsidized health insurance for the poor: evaluating the Colombian experience using propensity score matching." International journal of health care finance and economics 5, no. 3 (2005): 211-239.

Wamser, Georg. "The Impact of Thin‐Capitalization Rules on External Debt Usage–A Propensity Score Matching Approach." Oxford Bulletin of Economics and Statistics 76, no. 5 (2014): 764-781.

25

Table 1: Naïve Result (LPM)

Outcome Variables Coefficient Standard Error T-Statistic of IAY Pucca Wall .064 .010 5.96*** Pucca Roof .021 .009 2.16**

Pucca Floor .009 .009 1.02

Pucca House .009 .010 0.93

No Toilet .002 .010 0.22

Pit Toilet .016 .008 1.86*

Septic Toilet -.017 .008 -2.10**

Flush Toilet -.0006 .003 -0.18

Open Fire Chulha .021 .010 2.02**

Traditional Chulha -.031 .012 -2.64***

Improved Chulha .010 .007 1.42

Non-biomass Chulha .0007 .007 0.10

Adequate Availability of water -.0007 .006 -0.11

Tube Well Water -.021 .007 -2.90***

Handpump Water .067 .010 6.34***

Open Well Water -.019 .007 -2.62***

Covered Well Water -.009 .001 -6.49***

Piped Water -.012 .010 -1.12

Electricity -.037 .008 -4.20***

Excrement -.000 .011 -0.01

Stagnant Water .026 .010 2.57***

Village Health Facility .017 .012 1.42

26

Table-2: Descriptive Statistics of Covariates (Pre- & Post-Matching) & Logit Result of Participation

Variable Matching Status

Mean Treatment

Mean Control

% |Bias| Reduction

T-Stat Logit

I II III IV V VI BPL U

M .642 .648

.323

.649 99.7

30.69*** -0.05

.959*** (.055)

Lived in rural area in 2011=1

U M

.923

.924 .639 .939

94.9

27.19*** -1.69

1.317*** (.102)

Assets in 2005 (index)

U M

10.942 11.13

15.063 11.013

97.2

-30.42*** 0.70

-.029*** (.006)

Agricultural Land U M

.534

.553 .441 .602

48.6

8.54*** -2.85***

-.171*** (.056)

Education-male household head

U M

5.435 5.33

8.023 5.334

99.8

-22.57*** -0.02

-.025*** (.007)

Education-Female household head

U M

2.940 2.957

5.79 2.926

98.9

-24.37*** 0.23

-.042*** (.007)

Household size U M

4.960 5.206

4.847 5.218

89.8

2.21** -0.15

.045*** (.016)

SC/ST=1 U M

.526

.513 .285 .500

94.6

23.98*** 0.75

.550*** (.053)

Religion U M

1.250 1.247

1.322 1.262

80.4

-3.79*** -0.48

-.041 (.030)

Number of male adults

U M

1.375 1.487

1.430 1.472

72.4

-2.86*** 0.59

.047 (.048)

Number of female adults

U M

1.416 1.440

1.494 1.437

95.8

-4.50 0.14

-.021 (.054)

Age of male household head

U M

49.019 48.65

49.168 48.465

-24.5

-0.46 0.43

.013* (.007)

Age of female household head

U M

44.172 43.035

44.652 42.851

61.7

-1.61 0.45

-.011 (.007)

Member of political party =1

U M

.027

.029 .038 .029

94.8

-2.68*** 0.10

-.119 (.151)

Knows elected official=1

U M

.082

.087 .134 .090

94.4

-6.91*** -0.30

-.277*** (.091)

Attend public meeting =1

U M

.437

.458 .284 .439

87.3

15.20*** 1.14

.503*** (.053)

27

Confidence in Panchayat =1

U M

1.900 1.913

1.915 1.905

42.0

-1.00 0.38

.054*** (.037)

District U M

17.11 17.179

14.711 16.928

89.5 |

8.57 *** 0.52

.015*** (.001)

State U M

19.616 19.737

18.3 19.674

95.2

6.04*** 0.21

.011*** (.003)

N 33605 Pseudo R2 0.136

Table-3: Summary Measures of Matching Quality

Summary of Covariate Balance before and after Matching

Pseudo-R2

0.138

0.003

LR Chi2

1868.63

15.48

P> Chi2

0.000

0.692

Mean Bias 27.5 1.9

Median Bias 16.7 1.4

Figure-1: Graph of Common Support

28

Table-4 PSM Estimation Radius Matching (Caliper 0.01)

Radius Matching (0.01) Mean Mean Mean ATT Treated Control Difference S.E T-stat Pucca Wall .627 .545 .082 .012 6.73*** Pucca Roof .784 .753 .030 .010 2.89***

Pucca Floor .471 .443 .028 .012 2.22**

No Toilet .683 .690 -.007 .012 -0.66

Pit Toilet .146 .118 .028 .009 3.09***

Septic Toilet .151 .156 -.004 .009 -0.50

Flush Toilet .015 .029 -.014 .003 -3.70***

Open Fire Chulha .247 .228 .019 .010 1.79*

Traditional Chulha .521 .544 -.022 .012 -1.78*

Improved Chulha .100 .085 .014 .007 1.93*

Non-biomass Chulha .125 .134 -.009 .009 -1.08

Adequate Water Availability .924 .925 -.001 .006 -0.23

29

Tube Well Water .092 .114 -.021 .007 -2.92***

Handpump Water .421 .357 .064 .012 5.11***

Open Well Water .093 .115 -.021 .007 -2.95***

Covered Well Water .002 .010 -.008 .001 -5.06***

Piped Water .355 .362 -.007 .012 -0.57

Electricity .762 .786 -.024 .010 -2.25**

Excrement .306 .303 .002 .011 0.20

Stagnant Water .217 .185 .031 .010 3.03***

Village Health Facility .466 .449 .017 .012 1.37

Table 5: Sensitivity Analysis: Mantel-Haenszel Bounds

Outcome Gamma Q_mh+ Q_mh- p_mh+ p_mh-

Pucca Walls 1 10.1395 10.1395 0 0

Pucca Walls 1.5 18.453 2.19917 0 .013933

Pucca Walls 2 24.7574 3.31986 0 .00045

Pucca Walls 2.5 29.9746 7.67564 0 8.2e-15

Pucca Walls 3 34.5002 11.2907 0 0

Pucca Roof 1 8.13088 8.13088 2.2e-16 2.2e-16

Pucca Roof 1.5 15.1681 1.40611 0 .079846

Pucca Roof 2 20.5067 3.26061 0 .000556

30

Pucca Roof 2.5 24.9263 6.96355 0 1.7e-12

Pucca Roof 3 28.7609 10.0399 0 0

Pucca Floor 1 17.1024 17.1024 0 0

Pucca Floor 1.5 26.1946 8.62694 0 0

Pucca Floor 2 33.2152 2.81373 0 .002448

Pucca Floor 2.5 39.0924 1.60636 0 .054098

Pucca Floor 3 44.232 5.26896 0 6.9e-08

No Toilet 1 20.4052 20.4052 0 0

No Toilet 1.5 12.1889 29.3451 0 0

No Toilet 2 6.63595 36.3102 1.6e-11 0

No Toilet 2.5 2.41548 42.1701 .007857 0

No Toilet 3 .9576 47.3112 .169132 0

Table 5 (contd.): Sensitivity Analysis: Mantel-Haenszel Bounds


Pit Toilet 1 .214269 .214269 .415169 .415169

Pit Toilet 1.5 6.0406 5.52531 7.7e-10 1.6e-08

Pit Toilet 2 10.2948 9.76577 0 0

Pit Toilet 2.5 13.7149 13.1859 0 0

Pit Toilet 3 16.6113 16.098 0 0

Septic Toilet 1 14.1993 14.1993 0 0

Septic Toilet 1.5 21.0053 7.9058 0 1.3e-15

Septic Toilet 2 26.2649 3.63273 0 .00014

31

Septic Toilet 2.5 30.6523 .370712 0 .355426

Septic Toilet 3 34.4703 2.21729 0 .013302

Flush Toilet 1 11.7722 11.7722 0 0

Flush Toilet 1.5 15.1511 8.85306 0 0

Flush Toilet 2 17.8861 7.00259 0 1.3e-12

Flush Toilet 2.5 20.2285 5.66718 0 7.3e-09

Flush Toilet 3 22.2989 4.62857 0 1.8e-06

Open Fire Chulha 1 10.3737 10.3737 0 0

Open Fire Chulha 1.5 3.27032 17.8735 .000537 0

Open Fire Chulha 2 1.62461 23.6023 .052123 0

Open Fire Chulha 2.5 5.48331 28.3664 2.1e-08 0

Open Fire Chulha 3 8.67548 32.5129 0 0



Traditional Chulha 1 10.5582 10.5582 0 0

Traditional Chulha 1.5 2.31685 19.1763 .010256 0

Traditional Chulha 2 3.41427 25.6967 .00032 0

Traditional Chulha 2.5 7.93461 31.0811 1.1e-15 0

Traditional Chulha 3 11.689 35.744 0 0

Improved Chulha 1 4.91615 4.91615 4.4e-07 4.4e-07

Improved Chulha 1.5 .039873 9.98825 .484097 0

Improved Chulha 2 3.33126 13.819 .000432 0

32

Improved Chulha 2.5 6.05039 16.9809 7.2e-10 0

Improved Chulha 3 8.31934 19.7183 0 0

Non-Biomass Chulha 1 21.2575 21.2575 0 0

Non-Biomass Chulha 1.5 28.5059 14.7681 0 0

Non-Biomass Chulha 2 34.2371 10.5057 0 0

Non-Biomass Chulha 2.5 39.09 7.34102 0 1.1e-13

Non-Biomass Chulha 3 43.36 4.8203 0 7.2e-07

Adequate Water Availability 1 .324078 .324078 .372939 .372939

Adequate Water Availability 1.5 4.64545 3.9011 1.7e-06 .000048

Adequate Water Availability 2 7.81482 7.0473 2.8e-15 9.1e-13

Adequate Water Availability 2.5 10.3798 9.57473 0 0

Adequate Water Availability 3 12.5693 11.7152 0 0



Tube Well Water 1 2.77664 2.77664 .002746 .002746

Tube Well Water 1.5 7.63572 1.88457 1.1e-14 .029744

Tube Well Water 2 11.2444 5.28215 0 6.4e-08

Tube Well Water 2.5 14.179 7.98777 0 6.7e-16

Tube Well Water 3 16.6844 10.2697 10.2697 10.2697

Handpump Water 1 15.1068 15.1068 0 0

Handpump Water 1.5 6.83746 23.9315 4.0e-12 0

Handpump Water 2 1.13364 30.7236 .128473 0

33

Handpump Water 2.5 3.22166 36.3992 .000637 0

Handpump Water 3 6.84412 41.3562 3.8e-12 0

Open Well Water 1 1.38642 1.38642 .082809 .082809

Open Well Water 1.5 3.28538 6.19122 .000509 3.0e-10

Open Well Water 2 6.72805 9.7397 8.6e-12 0

Open Well Water 2.5 9.48012 12.6252 0 0

Open Well Water 3 11.8023 15.0967 0 0

Covered Well Water 1 4.36664 4.36664 6.3e-06 6.3e-06

Covered Well Water 1.5 5.69363 3.21818 6.2e-09 .000645

Covered Well Water 2 6.77159 2.48593 6.4e-12 .006461

Covered Well Water 2.5 7.70054 1.95407 6.8e-15 .025346

Covered Well Water 3 8.52757 1.53785 0 .062043



Piped Water 1 10.4801 10.4801 0 0

Piped Water 1.5 18.7862 2.54447 0 .005472

Piped Water 2 25.0656 2.9619 0 .001529

Piped Water 2.5 30.243 7.30489 0 1.4e-13

Piped Water 3 34.7193 10.9122 0 0

Electricity 1 14.9508 14.9508 0 0

Electricity 1.5 22.6736 7.80791 0 2.9e-15

Electricity 2 28.6801 2.9317 0 .001686

34

Electricity 2.5 33.7326 .750634 0 .226437

Electricity 3 38.1651 3.81394 0 .000068

Excrement 1 8.25239 8.25239 1.1e-16 1.1e-16

Excrement 1.5 .758179 16.0548 .224172 0

Excrement 2 4.47375 21.9471 3.8e-06 0

Excrement 2.5 8.62265 26.8104 0 0

Excrement 3 12.0753 31.0211 0 0

Stagnant Water 1 7.15475 7.15475 4.2e-13 4.2e-13

Stagnant Water 1.5 .466921 14.1168 .320278 0

Stagnant Water 2 4.19512 19.3764 .000014 0

Stagnant Water 2.5 7.90498 23.7187 1.3e-15 0

Stagnant Water 3 10.9939 27.479 0 0

Gamma : odds of differential assignment due to unobserved factors

Q_mh+ : Mantel-Haenszel statistic (assumption: overestimation of treatment effect)

Q_mh- : Mantel-Haenszel statistic (assumption: underestimation of treatment effect)

p_mh+ : significance level (assumption: overestimation of treatment effect)

p_mh- : significance level (assumption: underestimation of treatment effect)

Table 6: Inverse Probability Weighted Regression

Outcome Variables Coefficient Standard Error Z-Statistic of IAY Robust Pucca Wall .071 .010 6.56*** Pucca Roof .023 .009 2.41**

35

Pucca Floor .020 .009 2.18**

No Toilet -.012 .010 -1.22

Pit Toilet .034 .008 3.92***

Septic Toilet .002 .008 0.37

Flush Toilet -.011 .003 -3.6***

Open Fire Chulha .022 .010 2.01**

Traditional Chulha -.019 .012 -1.63

Improved Chulha .014 .007 1.92*

Non-biomass Chulha -.002 .006 -0.43

Adequate Availability of water -.002 .006 -0.32

Tube Well Water -.021 .007 -2.93***

Handpump Water .063 .010 5.97***

Open Well Water -.020 .007 -2.76***

Covered Well Water -.012 .000 -16.17***

Piped Water -.009 .010 -0.86

Electricity -.030 .008 -3.74***

Excrement -.000 .011 -0.04

Stagnant Water .032 .010 3.08***

Village Health Facility .019 .012 1.54

Table-7: Nearest Neighbor Matching

Nearest Neighbor Matching Mean Mean Mean

ATT Treated Control Difference S.E T-stat Pucca Wall .627 .543 .084 .013 6.23*** Pucca Roof .783 .754 .028 .011 2.49** Pucca Floor .471 .430 .041 .013 2.96***

No Toilet .682 .701 -.018 .013 -1.44

Pit Toilet .147 .117 .029 .009 3.04***

Septic Toilet .151 .147 .003 .010 0.39

Flush Toilet .015 .030 -.015 .003 -3.90***

Open Fire Chulha .248 .227 .020 .011 1.72*

Traditional Chulha .521 .546 -.025 .013 -1.80*

Improved Chulha .100 .087 .012 .008 1.50

36

Non-biomass Chulha .124 .132 -.007 .009 -0.84

Adequate Water Availability .924 .926 -.002 .007 -0.30

Tube Well Water .092 .114 -.022 .008 -2.75***

Handpump Water .421 .367 .053 .013 3.91***

Open Well Water .093 .115 -.021 .008 -2.57***

Covered Well Water .002 .009 -.006 .001 -3.90***

Piped Water .355 .352 .003 .013 0.23

Electricity .763 .798 -.035 .011 -3.00***

Excrement .306 .297 .009 .012 0.72

Stagnant Water .217 .184 .032 .011 2.88***

Village Health Facility .467 .435 .032 .013 2.30**

Documents

Indira Awas Yojana and Housing Adequacy: An Evaluation